feat: add prompt-type Stop hook for LLM completion verification

[mabry1985]

Summary

• Adds a type: prompt Stop hook that runs an LLM evaluation before every session end to catch incomplete work
• Detects four failure patterns: silent exit, partial work, error abandonment, and untested changes
• Chains before the existing evaluate-session.js pattern miner — verifier runs first, miner only fires if work is complete

Context

Inspired by patterns from claude-howto. The prompt-type hook uses an LLM to evaluate last_assistant_message rather than regex pattern matching, catching nuanced failure modes like agents that hit an error and suggest the user handle it instead of investigating.

Test plan

• CI passes
• Verify hook fires on session end (check for 30s prompt evaluation)
• Verify normal Q&A sessions pass through without blocking
• Verify incomplete work (e.g., planned edits not made) triggers continuation

🤖 Generated with Claude Code

Summary by CodeRabbit

• Chores
  • Enhanced internal verification mechanisms to improve system reliability and error detection.

[coderabbitai]

Caution

Review failed

Pull request was closed or merged during review

📝 Walkthrough

Walkthrough

A new Stop hook was added to .claude/settings.json that deploys a prompt-type completion verifier with a 30-second timeout. The hook evaluates the assistant's final message for four incomplete-work patterns and returns either a description of remaining work or confirmation that work is complete.

Changes

Cohort / File(s)	Summary
Completion Verification Hook .claude/settings.json	Added a Stop hook that runs a prompt-type completion verifier (30-second timeout) to check for incomplete work patterns: silent exit, partial work, error abandonment, and untested changes. Returns remaining work or "Work is complete."

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~5 minutes

Poem

🐰 A hook now watches work with care,
Checking if the task is fair,
Silent exits, partial mends—
Our verifier sees where work ends!
Complete or not, the truth it tells,
In .claude where the checker dwells. 🌟

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately describes the main change: adding a prompt-type Stop hook for LLM completion verification in .claude/settings.json. It is concise, specific, and clearly conveys the primary purpose of the changeset.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/stop-hook-completion-verifier

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}